Radio 3 Subjective Technical Quality

[David-G]

This explains that the proms were fed into BH at 48KHz sampling rate, down-sampled to 44.1 for the continuity suite, up-sampled to 48kHz for the MP2 encoding area, and finally down-sampled to 44.1Khz for the AAC feed.

I would imagine that this would involve a considerable loss of quality? Also - what is an SRC event?

[Bryn]

Sampling Rate Conversion?

[OldTechie]

I would imagine that this would involve a considerable loss of quality? Also - what is an SRC event?

Thanks Bryn - I forgot to make myself comprehensible.

Actually, the change between sampling rates is pretty low loss. There are various algorithms to do it and with both hardware and software implementations. Virtually all CDs are mastered at a higher sample rate and finally converted to 44.1 at a final stage. However, the reason that the BBC was in this up/down/up/down state was because they were trying to minimise the conversions in the most frequent operation. The normal standard in broadcasting is 48kHz sampling - but they had a storage system that was fixed at 44.1. They did not want a conversion every time something was stored or retrieved, so the continuity suite was running at the same rate as the storage system. Maybe it still is!

[Gordon]

Thanks for that OT. I will see if the little bird is still singing! My suspicion is that BH is now all 48 with CD being passed into the storage system via a SRC device - just the once, possibly in non real time. Similarly live CD could be injected using a real time SRC converter. Live music in studio goes into a digital desk. All externally produced material will probably be delivered as files in 48/24 and put straight into the storage system - any error between the external source 48 clock should be small enough to ignore, less than 1ppm say. In this case all the network feeds would originate in 48. It is worth noting that all the usual digital parameter tolerances apply - eg jitter on the master clock[s] at BH must be good if it is to be inaudible. We've done that one to death before too!! It needs to be maintained throughout the distribution network and therein lies a load of issues. What is the quality of the clock in the average computer sound card or DAB receiver?

SRC processing between 48/44.1 isn't trivial but it isn't that hard either, even in real time. One can always make a converter that is simple but it won't necessarily be very good. Good accuracy requires lots of taps to the digital filters but with diminishing returns. Unlike say 96 to 48, the ratio 48 to 44.1 is awkward - 480/441 = 160/147 = 8x20/7x21 = 1.0884353741496600.....

Synchronisation of the 44.1 player to the 48 system master is an issue in real time but pro CD players can be sync'd to external clocks so that the 44.1 used by the player could be derived from the master 48. The source clock in CD players is typically a crystal oscillator with a tolerance of about 1 ppm or better but they can be "pulled" away from their natural frequency in order to synchronise them with another clock. This ensures that there are no pitch issues in playback. Actually the CD master clock is at 4.31218 MHz, being 98 times the sampling frequency and 90.0375 times 48 kHz; sync'ing at 4.3218 might be better.

[OldTechie]

Thanks for that OT. I will see if the little bird is still singing!

Thanks Gordon, that would be nice to know.

If they have a new 48kHz sample rate storage system in service (and it could have fallen foul of the DMI fiasco) I would expect them to have given up playing CDs as far as possible and be ripping them directly and without SRC into the storage. I would expect them to have dumped the existing BWAV files from the old system into the new one, again without touching them so they would be 44.1. I would then expect the storage system to deliver station-synchronous 48kHz sample-rate data (and probably MADI.) That would make it much easier to track copyright on items played, rather than relying on someone writing down which CD they played!

I was looking at the kit they use at the sharp end for the proms (being impressed, yet again, at how good they are sounding to me.) They have a magic box for use in the hall. It has 153dB dynamic range to convert microphone signals to digital audio. So it can handle +22dBu on the input, and the thermal noise of the microphone is at the level of the least significant bit of the 28bit signal. You don't have the usual problem of setting the gain of a mic preamp to a suitable level for the following DAC - and if you can record the 28bit signal you can guarantee to capture an undistorted mic signal with no adjustments, even if you've stuffed a ball and biscuit mic in front of the Rolling Stones guitar amp! According to the SIS truck leaflet they are using this: http://www.stagetec.com/en/audio-tec...converter.html.

Of course measuring mic amp noise is always a bit of a matter of opinion - with input terminations of 600 ohms, 300 ohms, 200ohms, 150ohms and 60ohms being chosen by different people, and should it be 10kHz or 20kHz bandwidth. So whether -131dBu (150ohm/20kHz) is right for the lowest significant bit might be a matter of opinion.

Do you think they use some exotic microphone cable to the converter? I have a suspicion they probably use good old BBC PSF4/1 in one of its current Canford variants, in that avoiding thyristor buzz from the TV lights is probably quite high on the list of priorities. There are certainly some who go in for exotic microphone cables (http://www.channelclassics.com/equipment/) though I would question the acoustic improvement of painted wood-effect finish to the metal front cover of a rack mounted ADC.